To change sexual behavior for the purpose of AIDS prevention, social marketing programs balance the use of mass media with the use of interpersonal communications. Studies have shown that interpersonal communication, such as peer education and promotion of contraceptive methods by providers, can have a significant impact on contraceptive use.1 However, evidence about the magnitude and quality of impact that mass media campaigns have on promoting behavior change is less readily available, in part because of the difficulty of designing experimental studies to measure the impact of mass media campaigns.2
In the absence of experimental studies, it is possible to estimate the impact of mass media campaigns on condom use by using path analysis. Path analysis is a regression-based approach that uses cross-sectional data but permits a set of variables to be ordered in a manner consistent with causal interpretation.3 In this article, we use path analysis to assess whether a mass media campaign designed to promote use of the female condom had an impact on intentions to use the female condom among men and women of reproductive age in Tanzania. (Previous research has shown that the intention to use a contraceptive method is a powerful predictor of future use.4)
BACKGROUND
A substantial body of communications literature based on experience in developed countries has shown that mass media campaigns are not necessary or sufficient in themselves to produce changes in behavior. In fact, mass media messages usually reinforce attitudes; it is less common for them to completely change a person's opinion. The effect of mass media on behavior is indirect and operates through various factors. For example, how people communicate with each other about the mass media messages to which they are exposed may determine their response.5
In developing countries, analyses of cross-sectional surveys have shown strong associations between contraceptive use and exposure to mass media.6 Because selfselection may explain this association (i.e., those who are already convinced of the usefulness of contraception may be more likely to remember messages promoting it), cross-sectional associations cannot be interpreted as evidence of an impact of mass media campaigns.7 Instead, if we are to understand how and to what extent mass media affect behavior or its antecedents, it is important to identify plausible mechanisms through which mass media have an impact.
Diffusion theorists postulate that mass media affect contraceptive use by stimulating partners to discuss contraceptive use.8 Through the sharing of information and mutual feedback, people give meaning to information, understand each other's views and influence each other.9 Thus, discussion of contraceptive use leads to the development of a better understanding between partners of their reproductive health goals. Indeed, the perception that one's partner disapproves is an important deterrent to contraceptive adoption.10
When individuals' goals coincide or when they can negotiate agreement concerning the need for contraceptive use, couples are likely to adopt a method. Several studies have shown that discussing contraceptive use with a partner is highly predictive of future contraceptive adoption.11 One recent study showed that mass media campaigns can stimulate discussions of reproductive health issues,12 while another demonstrated that the indirect effects of interventions on behavioral antecedents have a powerful impact on contraceptive behavior.13
In this article, we assess whether mass media promotion of the female condom (via radio and newspaper) motivated Tanzanian men and women to use this method. We examine whether messages in the mass media stimulated discussion about female condom use and, in turn, whether such discussion influenced women's and men's intentions to use the female condom.
Interventions using mass media and interpersonal communication are particularly likely to have an impact on behavior.14 The intervention being evaluated had two components besides the mass media campaign—peer education, and explanation by medical providers of how to use the female condom. We also assess here the impact of these two interpersonal communication components on the intention to use the female condom.
THE FEMALE CONDOM INTERVENTION
Population Services International (PSI) introduced the female condom in Dar es Salaam toward the end of 1998.* PSI had experience in marketing the female condom through its programs in Zambia and Zimbabwe, and lessons learned in these countries were used to develop a marketing strategy for the female condom in Tanzania.15
A fundamental issue related to condom adoption in Sub-Saharan Africa is women's difficulty in negotiating the use of condoms.16 Because condoms carry the stigma of sexually transmitted infections (STIs), proposing condom use raises the possibility either that one does not trust one's partner or that one is already infected with an STI. To avoid stigmatizing the female condom, PSI marketed it as a contraceptive method with the added benefit of STI protection. Research in Kenya has shown that male partners are more accepting of the female condom when it is introduced as a family planning method, even if women actually use it for disease prevention.17
Messages that had been developed for the female condom intervention in Zimbabwe were pretested through focus groups with target audiences in Tanzania to ensure they were suitable for the Tanzanian social context. Consistent with messages in Zambia and Zimbabwe, the emphasis in advertising and promotion was to provide women with the language and tools they needed to discuss and negotiate use of the female condom. Because the intervention was intended to promote discussion of the female condom between partners, communications messages were targeted to both men and women. The product was marketed as a method for couples who wanted to protect themselves against pregnancy and HIV. It was branded as "care," with the caption "For couples who care."
A mass media campaign to promote the female condom was implemented during 1999. Radio messages promoting female condom use were aired in April-May and in October- November 1999. The print media were another potential source of information, so we placed advertisements for the "care" female condom in national newspapers. Our marketing efforts discussed the female condom as a family planning method to avoid the stigma of distrust and infidelity associated with the male condom. We expected that this approach would enable women to discuss the female condom with a partner without challenging social norms.
Focus groups were conducted with women approximately eight months after the launch of "care" to assess the effectiveness of the marketing strategy. These indicated that the strategy of emphasizing family planning and AIDS prevention messages had been effective: Participants associated use of the female condom with marital relationships, but reported that the most important reason for their own use of the female condom was HIV prevention.18
Approaches to interpersonal communication were integral components of the female condom intervention. Previous experience in Sub-Saharan Africa had shown that peer counseling improves women's ability to negotiate female condom use and that communications between couples increased the method's acceptability.19 Potential users were given a detailed explanation of female anatomy by community-based peer educators and health care workers, such as nurses, doctors and pharmacists, who were trained to counsel potential users.
Because the project was aimed at middle- and upper-income professionals for reasons of cost recovery, pharmacies were a logical primary distribution point for the female condom.† The female condom also was sold through nongovernmental organization (NGO) providers and community-based agents and was priced at US $0.44 for a pack of two condoms.‡
DATA AND METHODS
Methods
The objectives of the 1999 Tanzania Female Condom Consumer Profile Survey (TFCCPS-99) were to identify the level of exposure to the female condom mass-marketing intervention among men and women of reproductive age, to determine the level of female condom use and to examine respondents' intentions to use the female condom in the future.
On the basis of sales of the female condom, we estimated that the level of female condom use in the general population of reproductive age in Tanzania was below 1%. Because of this low level of use, we decided to conduct an exit survey at outlets that sold the female condom instead of conducting a large and expensive household survey. The methodology used in Tanzania was adapted from that of a similar survey conducted in Zambia.20
The survey was restricted to pharmacies and NGO facilities in Dar es Salaam where the female condom was being sold. Multistage random sampling was used to draw a representative sample of outlets. The first stage involved creating a master sampling list of outlets that sold the "care" female condom and drawing a stratified random sample that included 50% of all such outlets. Of the 67 outlets on the master list (58 pharmacies and nine NGOs), 33 (29 pharmacies and four NGOs) were selected. The probability that a pharmacy was selected was 0.50; for NGOs, it was 0.44. In the second stage, women and men who exited the outlets were randomly selected and invited to participate; only those aged 15-49 were eligible.
Using questionnaires used in Zambia and Zimbabwe as models, we prepared a quantitative questionnaire appropriate to the Tanzanian context for use in the TFCCPS-99.21 The questionnaire was designed to gather information on the respondents' social and demographic characteristics, as well as information on their knowledge of, discussion of, ever-use of and intention to use the female condom. This questionnaire was translated into Kiswahili and pretested before the survey was implemented.
Data Collection
Twenty-six interviewers (13 female and 13 male) were recruited and participated in a five-day training session during December 1999. Training included an explanation of the survey objectives and methodology. Although the interviewers were experienced, they were given refresher training on interviewing techniques. The questionnaire was discussed in detail, and interviewers became familiar with it by participating in role-plays. The final day of training was used to administer practice interviews in the field. One interviewer dropped out on the first day of data collection because she became ill; the remaining 12 female and 13 male interviewers collected the data from the last week of December 1999 through January 2000.
Interviewers were deployed at outlets in pairs. While one interviewed a client, the other counted men and women of reproductive age who entered the outlet. The number of clients entering the outlet on a particular day and the number of refusals were recorded to permit weighting of the sample. There were two shifts per day at each outlet. The first pair of interviewers (one of each sex) arrived at a provider outlet at 8:30 a.m. and stayed until 1:00 p.m. The second pair arrived at 1:00 p.m. and stayed until 6:00 p.m. Because of the different working hours at NGO facilities, only one shift was conducted at each of these outlets (from 8:30 a.m. to 2:30 p.m.)
A total of 3,029 men and women were interviewed. During data entry and cleaning, 16 records were eliminated because of missing data, leaving 3,013 respondents. The data were weighted to take into account the different probabilities of selection of outlet types and of individuals at each outlet, as well as the refusal rate at each outlet. During the 13 days of fieldwork at 29 pharmacies and four NGO clinics, 11,175 adult men and women (aged 15 or older) visited these outlets. The frequencies reported here are based on weighted numbers, but the numbers of cases shown are based on the unweighted numbers.
The Path Model and Variables
We use a simple model to assess the impact of mass media promotion on intentions to use the female condom in the future. Path analysis permits the assessment of both direct and indirect effects of independent variables on dependent variables. This technique is particularly useful in assessing the impact of mass media on motivation, because developed-country literature indicates that mass media has an indirect effect on motivation and because diffusion theory postulates that the impact of mass media on motivation operates by encouraging discussion between partners.
Exogenous variables included such social and demographic characteristics as age (in completed years), number of years of formal schooling (in completed years) and partnership status (single vs. other), as well as exposure to components of the social marketing intervention—peer education, explanation by a medical provider of how to use the female condom and radio or newspaper campaigns. We included two endogenous variables in the model—discussion of the female condom with a partner and intention to use the female condom in the future.
This model, which shows the hypothesized relationships between variables, appears in Figure 1 (page 153). Social and demographic characteristics of respondents (age, marital status and education) serve as controls in the path model. We are primarily interested in the impact of peer educators, provider promotion and the mass media campaign on intentions to use the female condom. In particular, we are interested in exploring whether the mass media have their effect through promoting discussion of the female condom between partners or through another, unidentified mechanism. We also want to assess the impact of peer education and provider promotion on the intention to use the female condom.
Statistical Analysis
We used the LISREL package for the analysis.22 We ran a two-group analysis using separate covariance matrices for men and women; this allowed us to test for differences in the causal pathways for men and women.
The final models shown in Figures 2 and 3 were arrived at by means of a three-step process. In the first step, we estimated a baseline model. In this model, we assumed:
• the applicability of the structural model shown in Figure 1, in which all exogenous variables (age, marital status, education, peer education, provider explanation and mass media messages) affect all endogenous variables (discussion of the female condom with the partner and intention to use the female condom), and discussion of the female condom with the partner also has an effect on the intention to use this method; and
• the causal models for men and women are identical.
The latter assumption was arrived at by forcing the unstandardized regression coefficients in the male and female subsamples to be equal. The fit for the baseline model was marginally acceptable.§
In the second step, we used modification indices to decide which equality constraints should be lifted. The fit for this model was very good and a significant improvement over that of the baseline model (X2[4]=39.0, p=.000).**
In the third and final step, we made the model more efficient by fixing nonsignificant regression coefficients at zero. This final model is presented in Figures 2 and 3, as its fit was not significantly worse than that of the Step 2 model (X2[6]=11.0, p=.089).†† All coefficients shown in these figures are significant at =.050. Correlations among the exogenous variables and residual effects are not shown in the figures. (Coefficients estimated under an equality constraint are indicated as such in the path diagrams.)
Sample Description
About 10% of respondents (301 respondents out of 3,013) were not sexually experienced. Because only sexually experienced persons were targeted for the female condom intervention, these 301 sexually inexperienced persons are not included in any of the subsequent analyses. All subsequent analyses are thus based on 2,712 respondents (1,186 women and 1,526 men). The mean age of respondents was 29 years, and they averaged 13.4 years of formal schooling. About 42% were single.
About 6% of respondents (8% of women and 5% of men, p<.01) reported that peer educators had discussed female condom use with them, and about 6% of respondents (9% of women and 5% of men, p<.01) said that medical providers had discussed the method. In contrast, radio and newspaper promotion had reached a much larger percentage of respondents: About 38% of respondents (41% of men and 34% of women, p<.01) reported having received information about the female condom through radio or newspapers.
Approximately 8% of respondents had discussed use of the female condom with a partner, 11% intended to use the device in the future and 3% had ever used it; these proportions did not differ by gender.
RESULTS
The models estimated and shown in Figures 2 and 3 predict respondents' likelihood of discussing the female condom with their partner and their intention to use the female condom. The explained variance in the discussion of the female condom with one's partner is around 10% for both men and women. The model explains almost twice as much of the variance in intention to use the female condom for the male sample (R2=.220) as for the female sample (R2=.117).
Women's Intentions
• Impact of the program. Figure 2 is the final model showing the pathways through which programmatic and social and demographic variables influence a woman's intention to use the female condom. All paths shown in the model are statistically significant at p<.05.
After controlling for all variables in the model, we found that receipt of female condom-related messages via the mass media did not directly affect a woman's intention to use the female condom; this is indicated by the lack of an arrow from the "mass media" variable to the "intention to use the female condom" variable in Figure 2. Mass media exposure did have a statistically significant impact on intentions, however, with a total effect‡‡ of 0.009 (p<.01) and a standardized total effect of 0.014, but operated through increasing discussion of the female condom between a woman and her partner. In turn, discussion of the female condom was a powerful predictor of a woman's intention to use the female condom. These findings are consistent with communications research showing that mass media have a gradual or indirect impact on behavior.23
Peer education had a relatively strong direct influence on a woman's intention to use the female condom, but did not affect whether she discussed the female condom with her partner. In contrast, exposure to a provider's explanation had a small direct effect on the intention to use the female condom and a substantial indirect effect of 0.070, or 58% of the total effect (p<.001), and a standardized indirect effect of 0.067, by encouraging the woman's discussion of the female condom with her partner. The impact of peer education and provider's explanation on intentions to use the female condom confirm the importance of interventions that use interpersonal communication to promote female condom use.
• Impact of social and demographic variables. We expected age to have a negative effect on intention to use the female condom, because older women tend to be more interested in avoiding pregnancy than in preventing STIs and because other reliable methods are available for family planning. However, our analysis showed no statistically significant effect of age on a woman's intention to use the female condom.
It may be that other demographic variables, such as marital status and education, explain the relationship between age and the intention to use the female condom. Consistent with this argument, single women were more likely than married women to intend to use the female condom. A woman's level of education did not have a direct effect on her intention to use the female condom, but it had a weak indirect effect on it, by increasing the likelihood that a woman would discuss the female condom with a partner (total effect=0.002, p<.001, standardized total effect=0.029). Overall, these findings are consistent with previous analyses of intentions to use the female condom.24
Men's Intentions
• Impact of the program. Figure 3 shows the pathways through which social and demographic and programmatic variables influence a man's intention to use the female condom. As in the previous model, all paths shown are statistically significant at p<.05.
First, the effect of discussing the female condom with one's partner on intention to use the female condom was significantly greater among men than among women (0.444 vs. 0.288). Thus, the indirect effects of programmatic and social and demographic variables on intention to use the female condom were larger among men than among women.
After controlling for other variables, we found no direct influence of mass media exposure on a man's intention to use the female condom. However, a man exposed to mass media messages about the female condom was significantly more likely than a man who had not been exposed to discuss use of the female condom with his partner, and discussion with a partner was predictive of a man's intention to use the female condom. These findings are consistent with those regarding the impact of mass media on a woman's intentions.
No significant gender differences were observed in the effect of mass media exposure on respondents' discussions of the female condom with a partner. Although the total effect of mass media exposure on intention to use the female condom among men (0.014, p<.01, standardized total effect=0.022) was somewhat larger than among women, the difference was not statistically significant.
Consistent with the findings for women, peer education had a powerful direct influence on men's intentions to use the female condom. The direct effect was even larger among men than among women. Peer education also encouraged a man to discuss use of the female condom with his partner. The total effect of peer education on a man's intention to use the female condom was 0.362 (p<.001, standardized total effect=0.249), of which 22% was indirect (0.079, p<.001, standardized indirect effect=0.055). No gender differences were observed in the effect of exposure to a provider's explanation on discussion of the female condom with a partner or on intentions to use it.
• Impact of social and demographic variables. Older men were more likely to intend to use the female condom than were younger men; this finding contrasts with the finding among women. The effect of marital status on intention to use the female condom was the same for men and women: Single respondents had a higher intention to use the female condom than did married respondents. Also, as was the case among women, marital status did not affect the likelihood that men would discuss the female condom with a partner. Finally, a man's education increased the likelihood that he would discuss the female condom with a partner, but there was no direct relationship between his education and his motivation to use the female condom. Again, no gender differences were observed in the effects of education.
DISCUSSION
Our primary objective in this article was to assess the impact of the mass media on men's and women's intentions to use the female condom, after implementation in Dar es Salaam of an intervention that included mass media promotion of the female condom, peer education about the female condom and providers' explanation of how to use the female condom. A secondary objective was to evaluate the impact of peer education and providers' explanation on intentions to use the female condom. We used a path analytic approach that allowed us to assess the impact of the different components of the social marketing program on men's and women's intentions to use the female condom.
Our finding that mass media exposure had a significant positive impact on intentions to use the female condom, even after we controlled for the effects of social and demographic variables, is consistent with the interpretation that mass media promotion of the female condom motivated sexual partners to discuss use of the female condom, and that discussion of condom use exerted a strong influence on their intention to use the female condom. This pattern was observed for both women and men. Because lack of discussion of condoms between partners can be a significant barrier to condom use, these findings imply that mass media promotion of condoms may have a positive impact on safer sex behavior.
Mass media were, however, only one of the program variables that had an impact on the discussion of the female condom with a partner. Among women, a provider's explanation had a powerful impact on the likelihood of discussing the female condom with a partner, while among men both peer education and a provider's explanation encouraged discussion of the female condom with a partner. These findings show that interpersonal interventions that encourage men and women to use condoms are in part successful because they increase discussion of condom use between partners.
We also found that an intervention promoting the discussion of the female condom with a partner may be particularly successful in motivating men to use condoms, because the effect of discussion on intentions was significantly greater among men. For these reasons, peer education had a particularly strong impact on male intentions to use condoms.
Unlike mass media promotion, which had an indirect effect on the intention to use the female condom, peer education had a powerful direct impact on women's and men's intentions to use the female condom. These findings are consistent with previous research showing that peer education interventions can promote safer sex behaviors and reinforce the continued use of peer education interventions in motivating individuals to adopt safer sex behaviors.25
The effects of the peer education and the provider component of the female condom program on intentions to use the method were larger than the impact of the mass media component. However, relatively few respondents were reached by either a peer educator or by a provider (6% each). In contrast, about 38% of respondents were reached by the mass media campaigns. Thus, although the impact of mass media messages on an individual's motivation to use a method may be less powerful than peer education or provider explanation, mass media are important because of their considerably greater reach.
Studies have shown that the combination of interpersonal interventions and mass media campaigns are likely to have a measurable impact on reproductive behavior.26 Our findings here are consistent with the results of this previous research and show that interpersonal and mass media interventions have independent effects.
One limitation of our research is that its findings can be generalized only to individuals of higher socioeconomic status who visit retail outlets such as pharmacies. Similar path analyses using representative samples of the general population (i.e., from household surveys) would be important to confirm our findings.
Our findings also suggest that it would be useful to compare the cost-effectiveness of different components of a social marketing program using path analysis. Such an analysis could help determine the optimal allocation of communication resources between the interpersonal and mass media components of an intervention. Finally, the results of this study support the strategy of including mass media promotion, peer education and provider explanation of condom use in programs designed to promote use of the female condom.